Y. D. Chong PH4401: Quantum Mechanics III
Chapter 5: Quantum Electrodynamics
This chapter gives an introduction to quantum electrodynamics, the quantum theory of the electromagnetic field and its interactions with electrons and other charged particles. We begin by formulating a quantum Hamiltonian for an electron in a classical electromag- netic field. Then we study how to quantize Maxwell’s equations, arriving at a quantum field theory in which the elementary excitations are photons—particles of light. The final step is to formulate a theory in which electrons and photons are treated on the same quantum mechanical footing, as excitations of underlying quantum fields. Along the way, we will see how relativity can be accomodated with quantum theory. Quantum electrodynamics is an extremely rich and intricate theory, and we will leave out many important topics. Interested readers are referred to are Dyson’s 1951 lecture notes on quantum electrodynamics [1], and Zee’s textbook Quantum Field Theory in a Nutshell [2].
I. QUANTIZATION OF THE LORENTZ FORCE LAW
A. Non-relativistic electrons in an electromagnetic field
Consider a non-relativistic charged particle in an electromagnetic field. As we are mainly interested in the physics of electrons interacting with electromagnetic fields, we henceforth take the electric charge of the particle to be −e, where e = 1.602 × 10−19 C is the elemen- tary charge. To describe particles with an arbitrary electric charge q, simply perform the substitution e → −q in the formulas you will subsequently encounter. We wish to formulate the Hamiltonian governing the quantum dynamics of such a particle, subject to two simplifying assumptions: (i) the particle has charge and mass but is otherwise “featureless” (i.e., we ignore the spin angular momentum and magnetic dipole moment that real electrons possess), and (ii) the electromagnetic field is treated as a classical field, meaning that the electric and magnetic fields are definite quantities rather than operators. (We will see how to go beyond these simplifications later.) Classically, the electromagnetic field acts on the particle via the Lorentz force law, F(r, t) = −e E(r, t) + r˙ × B(r, t) , (5.1) where r and r˙ denote the position and velocity of the particle, t is the time, and E and B are the electric and magnetic fields. If no other forces are present, Newton’s second law yields the equation of motion m¨r = −e E(r, t) + r˙ × B(r, t) , (5.2) where m is the particle’s mass. To quantize this, we must first convert the equation of motion into the form of Hamilton’s equations of motion. Let us introduce the electromagnetic scalar and vector potentials Φ(r, t) and A(r, t): ∂A E(r, t) = −∇Φ(r, t) − , (5.3) ∂t B(r, t) = ∇ × A(r, t). (5.4)
86 Y. D. Chong PH4401: Quantum Mechanics III
We now postulate that the equation of motion (5.2) can be described by the Lagrangian
1 h i L(r, r˙, t) = mr˙ 2 + e Φ(r, t) − r˙ · A(r, t) . (5.5) 2
This follows the usual prescription for the Lagrangian as kinetic energy minus potential energy, with −eΦ serving as the potential energy function, except for the −er˙ · A term. To see if this Lagrangian works, plug it into the Euler-Lagrange equations ∂L d ∂L = . (5.6) ∂ri dt ∂r˙i The partial derivatives of the Lagrangian are: ∂L h i = e ∂ Φ − r˙ ∂ A ∂r i j i j i (5.7) ∂L = mr˙i − eAi. ∂r˙i
Now we want to take the total time derivative of ∂L/∂r˙i. In doing so, note that the A field has its own t-dependence, as well as varying with the particle’s t-dependent position. Thus, d ∂L d = mr¨i − e Ai(r(t), t) dt ∂r˙i dt (5.8) = mr¨i − e ∂tAi − e r˙j∂jAi.
(In the above equations, ∂i ≡ ∂/∂ri, where ri is the i-th component of the position vector, while ∂t ≡ ∂/∂t.) Plugging these expressions into the Euler-Lagrange equations (5.6) gives h i mr¨i = −e − ∂iΦ − ∂tAi +r ˙j ∂iAj − ∂jAi h i (5.9) ˙ = −e Ei(r, t) + r × B(r, t) i .
(The last step can be derived by expressing the cross product using the Levi-Cevita symbol, and using the identity εijkεlmk = δilδjm − δimδjl.) This exactly matches Eq. (5.2), as desired. We now use the Lagrangian to derive the Hamiltonian. The canonical momentum is ∂L pi = = mr˙i − eAi. (5.10) ∂r˙i The Hamiltonian is defined as H(r, p) = p · r˙ − L. Using Eq. (5.10), we express it in terms of p rather than r˙:
p + eA |p + eA|2 e H = p · − + eΦ − (p + eA) · A m 2m m (5.11) |p + eA|2 e |p + eA|2 e = − A · (p + eA) − + eΦ − (p + eA) · A . m m 2m m
87 Y. D. Chong PH4401: Quantum Mechanics III
After cancelling various terms, we obtain |p + eA(r, t)|2 H = − eΦ(r, t). (5.12) 2m This looks a lot like the Hamiltonian for a non-relativistic particle in a scalar potential, |p|2 H = + V (r, t). 2m In Eq. (5.12), the −eΦ term acts like a potential energy, which is no surprise. More inter- estingly, the vector potential appears via the substitution p → p + eA(r, t). (5.13)
What does this mean? Think about what “momentum” means for a charged particle in an electromagnetic field. Noether’s theorem states that each symmetry of a system (whether classical or quantum) is associated with a conservation law. Momentum is the quantity conserved when the system is symmetric under spatial translations. One of Hamilton’s equations states that dp ∂H i = , dt ∂ri which implies that if H is r-independent, then dp/dt = 0. But when the electromagnetic potentials are r-independent, the quantity mr˙ (which we usually call momentum) is not necessarily conserved! Take the potentials Φ(r, t) = 0, A(r, t) = Ctz,ˆ (5.14) where C is some constant. These potentials are r-independent, but the vector potential is time-dependent, so the −A˙ term in Eq. (5.85) gives a non-vanishing electric field: E(r, t) = −Cz,ˆ B(r, t) = 0. (5.15) The Lorentz force law then says that d (mr˙) = eCz,ˆ (5.16) dt and thus mr˙ is not conserved. On the other hand, the quantity p = mr˙ − eA is conserved: d (mr˙ − eA) = eCzˆ − eCzˆ = 0. (5.17) dt Hence, this is the appropriate canonical momentum for a particle in an electromagnetic field. We are now ready to go from classical to quantum mechanics. Replace r with the position operator ˆr, and p with the momentum operator pˆ. The resulting quantum Hamiltonian is
|pˆ + eA(ˆr, t)|2 Hˆ (t) = − eΦ(ˆr, t). (5.18) 2m
Note: the momentum operator is pˆ = −i~∇ in the wavefunction representation, as usual.
88 Y. D. Chong PH4401: Quantum Mechanics III
B. Gauge symmetry
The Hamiltonian (5.18) possesses a subtle property known as gauge symmetry. Sup- pose we modify the scalar and vector potentials via the substitutions Φ(r, t) → Φ(r, t) − Λ(˙ r, t) (5.19) A(r, t) → A(r, t) + ∇Λ(r, t), (5.20) where Λ(r, t) is an arbitrary scalar field called a gauge field. This is the gauge transfor- mation of classical electromagnetism, which as we know leaves the electric and magnetic fields unchanged. When applied to the Hamiltonian (5.18), it generates a new Hamiltonian |pˆ + eA(ˆr, t) + e∇Λ(ˆr, t)|2 Hˆ (t) = − eΦ(ˆr, t) + eΛ(˙ ˆr, t). (5.21) Λ 2m Now suppose ψ(r, t) is a wavefunction obeying the Schr¨odingerequation for the original Hamiltonian Hˆ : ∂ψ |pˆ + eA(ˆr, t)|2 i = Hˆ (t)ψ(r, t) = − eΦ(ˆr, t) ψ(r, t). (5.22) ~ ∂t 2m
Then it can be shown that the wavefunction ψ exp(−ieΛ/~) automatically satisfies the ˆ Schr¨odingerequation for the transformed Hamiltonian HΛ: ∂ ieΛ(r, t) ˆ ieΛ(r, t) i~ ψ(r, t) exp − = HΛ(t) ψ(r, t) exp − . (5.23) ∂t ~ ~ To prove this, observe how time and space derivatives act on the new wavefunction: ∂ ieΛ ∂ψ ie ieΛ ψ exp − = − Λ˙ ψ exp ∂t ∂t ~ ~ ~ (5.24) ieΛ ie ieΛ ∇ ψ exp − = ∇ψ − ∇Λ ψ exp . ~ ~ ~ When the extra terms generated by the exp(ieΛ/~) factor are slotted into the Schr¨odinger equation, they cancel the gauge terms in the scalar and vector potentials. For example, ieΛ h i ieΛ − i~∇ + eA + e∇Λ ψ exp − = (−i~∇ + eA) ψ exp − (5.25) ~ ~ If we apply the (−i~∇ + eA + e∇Λ) operator a second time, it has a similar effect but with the quantity in square brackets on the right-hand side of (5.25) taking the place of ψ:
2 ieΛ h 2 i ieΛ − i~∇ + eA + e∇Λ ψ exp − = |−i~∇ + eA| ψ exp − . (5.26) ~ ~ The remainder of the proof for Eq. (5.23) can be carried out straightforwardly. The above result can be stated in a simpler form if the electromagnetic fields are static. In this case, the time-independent electromagnetic Hamiltonian is |pˆ + eA(ˆr)|2 Hˆ = − eΦ(ˆr). (5.27) 2m
89 Y. D. Chong PH4401: Quantum Mechanics III
ˆ Suppose H has eigenenergies {Em} and energy eigenfunctions {ψm(r)}. Then the gauge- transformed Hamiltonian |pˆ + eA(ˆr) + e∇Λ(r)|2 Hˆ = − eΦ(ˆr) (5.28) Λ 2m has the same energy spectrum {Em}, with eigenfunctions { ψm(r) exp[−ieΛ(r)/~] }.
C. The Aharonov-Bohm effect
In quantum electrodynamics, it is the electromagnetic scalar and vector potentials that appear directly in the Hamiltonian, not the electric and magnetic fields. This has profound consequences. For example, even if a charged quantum particle resides in a region with zero magnetic field, it can feel the effect of nonzero vector potentials produced by magnetic fluxes elsewhere in space, a phenomenon called the Aharonov-Bohm effect. A simple setting for observing the Aharonov-Bohm effect is shown in the figure below. A particle is trapped in a ring-shaped region (an “annulus”), of radius R and width d R. Outside the annulus, we set −eΦ → ∞ so that the wavefunction vanishes; inside the annulus, we set Φ = 0. We ignore the z-dependence of all fields and wavefunctions, so that the problem is two-dimensional. We define polar coordinates (r, φ) with the origin at the ring’s center.
Now, suppose we thread magnetic flux (e.g., using a solenoid) through the origin, which lies in the region enclosed by the annulus. This flux can be described via the vector potential Φ A(r, φ) = B e , (5.29) 2πr φ
where eφ is the unit vector pointing in the azimuthal direction. We can verify from Eq. (5.29) that the total magnetic flux through any loop of radius r enclosing the origin is (ΦB/2πr)(2πr) = ΦB. The fact that this is independent of r implies that the magnetic flux density is concentrated in an infintesimal area surrounding the origin, and zero every- where else. However, the vector potential A is nonzero everywhere. The time-independent Schr¨odingerequation is
2 1 eΦB −i~∇ + eφ ψ(r, φ) = Eψ(r, φ), (5.30) 2m 2πr
90 Y. D. Chong PH4401: Quantum Mechanics III
with the boundary conditions ψ(R ± d/2, 0) = 0. For sufficiently large R, we can guess that the eigenfunctions have the form ( ψ cos π (r − R) eikRφ, r ∈ [R − d/2,R + d/2] ψ(r, φ) ≈ 0 d (5.31) 0 otherwise.
This describes a “waveguide mode” with a half-wavelength wave profile in the r direction (so as to vanish at r = R ± d/2), traveling in the azimuthal direction with wavenumber k. The normalization constant ψ0 is unimportant. We need the wavefunction to be single-valued under a 2π variation in the azimuthal coordinate, so n k · 2πR = 2πn ⇒ k = , where n ∈ . (5.32) R Z Plugging this into Eq. (5.30) yields the energy levels " # 1 n eΦ 2 π 2 E = ~ + B + ~ (5.33) n 2m R 2πR d e2 nh2 π2 2 = Φ + + ~ . (5.34) 8π2mR2 B e 2md2
These energy levels are sketched versus the magnetic flux ΦB in the figure below:
Each energy level has a quadratic dependence on ΦB. Variations in ΦB affect the energy levels despite the fact that B = 0 in the annular region where the electron resides. This is a manifestation of the Aharonov-Bohm effect.
It is noteworthy that the curves of different n are centered at different values of ΦB corresponding to multiples of h/e = 4.13567 × 10−5 T m2, a fundamental unit of magnetic flux called the magnetic flux quantum. In other words, changing ΦB by an exact multiple of h/e leaves the energy spectrum unchanged! This invariance property, which does not depend on the width of the annulus or any other geometrical parameters of the system, can be explained using gauge symmetry. When an extra flux of nh/e (where n ∈ Z) is
91 Y. D. Chong PH4401: Quantum Mechanics III
threaded through the annulus, Eq. (5.29) tells us that the change in vector potential is ∆A = (n~/er)eφ. But we can undo the effects of this via the gauge field ( n ∇Λ = −(n /er) e Λ(r, φ) = − ~ φ ⇒ ~ φ (5.35) e e−ieΛ/~ = einφ.
Note that this Λ is not single-valued, but that’s not a problem! Both ∇Λ and the phase factor exp(−ieΛ/~) are single-valued, and those are the quantities that enter into the gauge symmetry relations (5.19)–(5.20).
II. DIRAC’S THEORY OF THE ELECTRON
A. The Dirac Hamiltonian
So far, we have been using p2/2m-type Hamiltonians, which are limited to describing non-relativistic particles. In 1928, Paul Dirac formulated a Hamiltonian that can describe electrons moving close to the speed of light, thus successfully combining quantum theory with special relativity. Another triumph of Dirac’s theory is that it accurately predicts the magnetic moment of the electron. Dirac’s theory begins from the time-dependent Schr¨odingerwave equation, ˆ i~ ∂t ψ(r, t) = Hψ(r, t). (5.36)
Note that the left side has a first-order time derivative. On the right, the Hamiltonian Hˆ contains spatial derivatives in the form of momentum operators. We know that time and space derivatives of wavefunctions are related to energy and momentum by
i~ ∂t ↔ E, −i~ ∂j ↔ pj. (5.37) We also know that the energy and momentum of a relativistic particle are related by
3 2 2 4 X 2 2 E = m c + pj c , (5.38) j=1 where m is the rest mass and c is the speed of light. Note that E and p appear to the same order in this equation. (Following the usual practice in relativity theory, we use Roman indices j ∈ {1, 2, 3} for the spatial coordinates {x, y, z}.) Since the left side of the Schr¨odingerequation (5.36) has a first-order time derivative, a relativistic Hamiltonian should involve first-order spatial derivatives. So we make the guess
3 ˆ 2 X H = α0mc + αjpˆjc, (5.39) j=1
2 wherep ˆj ≡ −i~∂/∂xj. The mc and c factors are placed for later convenience. We now need to determine the dimensionless “coefficients” α0, α1, α2, and α3.
92 Y. D. Chong PH4401: Quantum Mechanics III
For a wavefunction with definite momentum p and energy E,
3 ! ˆ 2 X Hψ = Eψ ⇒ α0mc + αjpjc ψ = E ψ. (5.40) j=1
This is obtained by replacing thep ˆj operators with definite numbers. If ψ is a scalar, this 2 P would imply that α0mc + j αjpjc = E for certain scalar coefficients {α0, . . . , α3}, which does not match the relativistic energy-mass-momentum relation (5.38). But we can get things to work if ψ(r, t) is a multi-component wavefunction, rather than a scalar wavefunction, and the α’s are matrices acting on those components via the matrix- vector product operation. In that case,
3 ˆ 2 X H =α ˆ0mc + αˆjpˆjc, wherep ˆj ≡ −i~ ∂j, (5.41) j=1
where the hats on {αˆ0,..., αˆ3} indicate that they are matrix-valued. Applying the Hamil- tonian twice gives 3 !2 2 X 2 αˆ0mc + αˆjpjc ψ = E ψ. (5.42) j=1 This can be satisfied if 3 !2 2 X 2 ˆ αˆ0mc + αˆjpjc = E I, (5.43) j=1 where Iˆ is the identity matrix. Expanding the square (and taking care of the fact that the αˆµ matrices need not commute) yields
2 2 4 X 3 X 2 2 ˆ αˆ0m c + (ˆα0αˆj +α ˆjαˆ0) mc pj + αˆjαˆj0 pjpj0 c = E I. (5.44) j jj0
This reduces to Eq. (5.38) if theα ˆµ matrices satisfy
αˆ2 = Iˆ for µ = 0, 1, 2, 3, and µ (5.45) αˆµαˆν +α ˆναˆµ = 0 for µ 6= ν.
(We use Greek symbols for indices ranging over the four spacetime coordinates {0, 1, 2, 3}.) The above can be written more concisely using the anticommutator:
{αˆµ, αˆν} = 2δµν, for µ, ν = 0, 1, 2, 3. (5.46) ˆ Also, we need theα ˆµ matrices to be Hermitian, so that H is Hermitian. It turns out that the smallest possible Hermitian matrices that can satisfy Eq. (5.46) are 4 × 4 matrices. The choice of matrices (or “representation”) is not uniquely determined.
93 Y. D. Chong PH4401: Quantum Mechanics III
One particularly useful choice is called the Dirac representation: ˆ I 0ˆ 0ˆ σ ˆ1 αˆ0 = , αˆ1 = 0ˆ −Iˆ σˆ1 0ˆ (5.47) 0ˆ σ ˆ2 0ˆ σ ˆ3 αˆ2 = , αˆ3 = , σˆ2 0ˆ σˆ3 0ˆ where {σˆ1, σˆ2, σˆ3} denote the usual Pauli matrices. Since theα ˆµ’s are 4 × 4 matrices, it follows that ψ(r) is a four-component field.
B. Eigenstates of the Dirac Hamiltonian
According to Eq. (5.38), the energy eigenvalues of the Dirac Hamiltonian are s 2 4 X 2 2 E = ± m c + pj c . (5.48) j
This is plotted below:
The energy spectrum forms two hyperbolic bands. For each p, there are two degenerate positive energy eigenvalues, and two degenerate negative energy eigenvalues, for a total of four eigenvalues (matching the number of wavefunction components). The upper band matches the dispersion relation for a massive relativistic particle, as desired. But what about the negative-energy band? Who ordered that? It might be possible for us to ignore the existence of the negative-energy states, if we only ever consider an isolated electron; we could just declare the positive-energy states to be the ones we are interested in, and ignore the others. However, the problem becomes hard to dismiss once we let the electron interact with another system, such as the electromagnetic field. Under such circumstances, the availability of negative-energy states extending down to E → −∞ would destabilize the positive-energy electron states, since the electron can repeatedly hop to states with ever more negative energies by shedding energy (e.g., by emitting photons). This is obviously problematic. However, let us wait for a while (till SectionIID) to discuss how the stability problem might be resolved. For now, let us take a closer look at the meaning of the Dirac wavefunction. Its four components represent a four-fold “internal” degree of freedom, distinct from the electron’s
94 Y. D. Chong PH4401: Quantum Mechanics III ordinary kinematic degrees of freedom. Since there are two energy bands, the assignment of an electron to the upper or lower band (or some superposition thereof) consitutes two degrees of freedom. Each band must then posssess a two-fold degree of freedom (so that 2 × 2 = 4), which turns out to be associated with the electron’s spin.
To see explicitly how this works, let us pick a representation for theα ˆµ matrices. The choice of representation determines how the four degrees of freedom are encoded in the individual wavefunction components. We will use the Dirac representation (5.47). In this case, it is convenient to divide the components into upper and lower parts, ψ (r, t) ψ(r, t) = A , (5.49) ψB(r, t) where ψA and ψB have two components each. Then, for an eigenstate with energy E and momentum p, applying (5.47) to the Dirac equation (5.41) gives 1 X ψ = σˆ p ψ , (5.50) A E − mc2 j j B j 1 X ψ = σˆ p ψ . (5.51) B E + mc2 j j A j Consider the non-relativistic limit, |p| → 0, for which E approaches either mc2 or −mc2. For 2 the upper band (E & mc ), the vanishing of the denominator in Eq. (5.50) tells us that the 2 wavefunction is dominated by ψA. Conversely, for the lower band (E . −mc ), Eq. (5.51) tells us that the wavefunction is dominated by ψB. We can thus associate the upper (A) and lower (B) components with the band degree of freedom. Note, however, that this is only an approximate association that holds in the non-relativistic limit! In the relativistic regime, upper-band states can have non-vanishing values in the B components, and vice versa. (There does exist a way to make the upper/lower spinor components correspond rigorously to positive/negative energies, but this requires a more complicated representation than the Dirac representation [3].)
C. Dirac electrons in an electromagnetic field
To continue pursuing our objective of interpreting the Dirac wavefunction, we must de- termine how the electron interacts with an electromagnetic field. We introduce electromag- netism by following the same procedure as in the non-relativistic theory (SectionIA): add −eΦ(r, t) as a scalar potential function, and add the vector potential via the substitution pˆ → pˆ + eA(ˆr, t). (5.52) Applying this recipe to the Dirac Hamiltonian (5.41) yields ( ) 2 X h i i~ ∂tψ = αˆ0mc − eΦ(r, t) + αˆj − i~ ∂j + eAj(r, t) c ψ(r, t). (5.53) j You can check that this has the same gauge symmetry properties as the non-relativistic theory discussed in SectionIB.
95 Y. D. Chong PH4401: Quantum Mechanics III
In the Dirac representation (5.47), Eq. (5.53) reduces to